32 MP 4/3" sensor - why not?

[DriftForge]

No reason for there to be any less DR than the best P&S type cameras. Admittedly this may not be as high as people are used to in a DSLR, but I suspect you could still be looking at 6-7 stops, rather than 8-9.

 

[Jay Turberville]

No reason for there to be any less DR than the best P&S type cameras.
Admittedly this may not be as high as people are used to in a DSLR,
but I suspect you could still be looking at 6-7 stops, rather than
8-9.

Actually, the 2/3" 8Mp Sony sensor gives up one whole stop of dynamic range to the 8Mp 4/3" Kodak sensor in the E-300/500.

Now keep in mind, that this is dynamic range as determined by noise thresholds. Given the same degree of enlargement, a 32Mp image will be show the pixel noise at a much smaller size. So the perception of what is an acceptable noise threshold may be significantly different. In the end, you may not lose any practical dynamic range at all.

--
Jay Turberville
http://www.jayandwanda.com

 

[Joesiv]

How do you know that Olympus is using a "fairly aggressive AA filter?"

Maybe not all Oly cams do, but in the L10 review the per pixel sharpness in RAW with the 50mm ZD lens, the L10's sensor was much more sharp than the e-510. dpreview thought that it was likely the AA filter, and I would agree.
http://www.dpreview.com/reviews/panasonicdmcl10/page25.asp
--
Cloverdale, B.C., Canada
Olympus e-510
http://joesiv.smugmug.com

 

[BJL]

If a 32MP sensor distinctly outresolves all lenses for the 4/3 system, there would be no point to it, except marketing hype.

I disagree. The point would be to oversample the lens image instead of having to use an AA filter.

That might make sense, but maybe a weak AA filter and somewhat less pixels is a better way to get about the same results.

No, the fact that Olympus is still using a fairly agressive AA filter on its 10Mp sensors tells you that 10Mp isn't the limit.

I agree that 32MP might make sense with the better of 4/3 lenses, and I certainly did not want to suggest that 4/3 is already at its limit with the current 10MP. In fact it seems that some 4/3 lenses, including some zooms, can usefully go significantly beyond 10MP.

I was only saying that the ability of 4/3 lenses to handle the 2.7 micron pixel spacing of a 32MP 4/3 sensor is not proven by the fact that far smaller format digicam lenses can keep up with that pixel spacing.

--
I want all my lenses to be f/4 or brighter.

Bigger pixels are useless if they mean I need longer, slower telephoto lenses to get the same resolution with the same weight and cost.

 

[RedFox88]

well not 4/3 sensors, but they've made aps-c and aps-h 50MP prototype sensors last year. The news article said uses, if put into any products, might be security cameras to grab fine details that video tape would have no way of recording.

 

[Martin Datzinger]

RGBW - where W is unfiltered/pure luma with 2 stops higher sensitivity. And the sensor should then have a > 8fps 12MP no-bayer mode (with 60fps progressive 1920x1080 pixel-perfect still image quality-like H.264 output as a crop from a 2x2pixel binned 2000x1500 sensor readout) and 2 stops DR advantage in bayer and no-bayer (1fps) modes.



Of course, until then, they must get single RGB pixel's DR to at least today's D300 / 40D levels, for the white pixel to give any advantage. I have the impression, some active sensor cooling is necessary for that.

If they get fancy, they could also expand the 4/3 sensor in width to a 3:2 aspect ratio, which would allow a 16:9 crop with only very little vignetting and the same video output with less pixel density (if my calculations are correct 38MP bayer / 9,5MP non-bayer) and the same diagonal FOV, one is used with the 4/3 lenses at any FL setting:



Of course, the bayer conversion and H.264 (or then, any other compression) and data storage could also be done via a dedicated, less bulk- and heat-limited ASIC in an external video module, sporting amplified XLR mic connectors and additional battery slots, connected through 46Gb/sec fibre channel (assuming 16-bit sensor readout).

Kind regards,
Martin

--
http://www.datzinger.net

 

[Louis_Dobson]

1) You would lose a stop, tops.

2) Unless you pull and push stuff in RAW, DR is totally irrelevant anyway - the camera would still have way more than the current fashion for punchy pictures could allow onto paper or screen unless you use freak (and horrid) tone curves.
--
http://www.flickr.com/photos/acam

 

[Frank Dernie]

my first digital SLR was a Canon D30. It had an aps size sensor and 3.3 megapixels. I have excellent A4 prints I made from photographs I took with it.

10 megapixels is as much as I need to produce prints at A3 size. I do now own a Canon EOS 1Dsmk2 which has 16.7 megapixel 36x24 sensor, I use it where once I used a Rolleiflex 6008i. For circumstances where I used to use a 35mm film camera the E3 is better than film. I see no reason for more pixels, and in fact am heartily fed up that this and high iso performance seem to be the only thing internet fora discuss.

I would like more dynamic range and nicer gradation from existing resolution before a sensor with resolution enough for a 6 foot wide print I will never make but blown out highlights and blotchy gradation.
Just MHO

Frank

 

[esco]

I also get tired when I hear people saying that the current resolution and iso performance of the average dslr is enough. I too find the MP race of compact cameras pretty comical but for anyone that has looked at a 100% crop of a photo taken with a digital hassleblad setup and thier's no denying that there are still improvements to be made and that we should always welcome the evolution of technology.

Digital MF has actually taken off during the recent years and well I don't know about olympus but Canikon wants a piece of that pie very badly and without going beyond slr bodies, their plan is to make slr's so close in performance that MF becomes redundant. Yes laughable but certainly possible, we do have a 10fps dslr and a 21mp out today and unlike point and shoot cameras dslr's are nowhere near thier maturity.

--
Oldschool Evolt shooter

 

[Frank Dernie]

Yes but how big are your prints? If you only look at 100% crops on a computer monitor there is no point in taking a picture at all. Unless you print large posters you don't need 32 megapixels.

I have been shooting with my E3 today. None of the pictures have insufficient resolution for a sharp A4 print but several are clipped in both highlights AND shadows. More dynamic range would have made these technically better pictures, more pixels wouldn't.
Frank

 

[DriftForge]

Interesting value judgement there to suggest that prints have more value than a digitally displayed image. May be true, but to suggest that digital images have no value until printed... will leave that philosophical discussion for another day.

That aside 32MP works out to very close to 300 dpi on A3 and 150 dpi on A1. Whether or not that is of use for you, I suspect there are a commercially significant number of people for whom that level of detail would suit just fine.

 

[Kyle Jones]

..., better pixels instead of more of them. If a Foveon type sensor can deliver the same resolution as a Bayer sensor with 2x the pixels then the pixel numbers to get a 24 MP Bayer equivalent are a lot more reasonable. I think 32 MP equivalent would still put the diffraction limit in your face too early for the current lens line.
--
Street: http://www.wonderworks.com/streetphotographydigest.html
Flickr: http://www.flickr.com/photos/kyle_jones/

 

[Kyle Jones]

The idea wouldn't be to get the "full resolution" of the sensor. The
idea would be to eliminate the AA filter (which reduces contrast in
fine detail) and have a sensor that can record the "full resolution"
of the lens.

OK, I see where you guys are coming from now.
--
Street: http://www.wonderworks.com/streetphotographydigest.html
Flickr: http://www.flickr.com/photos/kyle_jones/

 

[munro harrap]

I wondered this often too. It makes sense. THe D300 Nikon DOES have shadow noise at 200, like the 8080, so why not. Why not in fact use that size pixel on a Full-frame 24x36mm sensor. The Nikon D3 will crop to 5:4 in-camera, and there are any number of cams you can alter format on, AND change lenses on, so why not?

Well, it'll all come together at Judgement Day, maybe. Right now the very sharpest lens can barely manage 2000 resolution edge to edge and corner to corner, so whilst there ARE lots of lenses that go to a MAX of 2300-2400 at certain apertures in the centre of the image, 10 megapixel in APS-C format is still optimistic overall.

A Zeiss Macro, and a few recent Nikkors are worth the upgrade, maybe, maybe. I think wait and the great DAY will come!!

Sooner or later the noise-free sensor will meet and wed the perfect optic and a new master race will be born! Until then, lets just stick with what works best... and lots of pixels AND lots of noise AND inferior optics do not IMHO appear to be the solution

 

[Ehrik]

That sounds interesting. I've also been thinking about W sensels.
And Kodak now has a patent for a layout with some W sensels too.
Of course, there must be reasons for the RGGB layout to be so
popular today, so there are probably drawbacks with your scheme.

To begin with, I've read that the commonly used colour filters lose
about one stop. They definitely don't lose two stops; that's 1/4 of
the light and we only need to split it into three colours.
So the white sensels can't be two stops more sensitive unless you
throw away an extra stop of light from the RGB sensels by using denser
filters, and I don't think we want to do that.

The white pixel will clip sooner than the RGB ones, so we'll lose
1/4 of the info of the highlights. The photon capture capacity per
area will go down, which will diminish the DR.

Also at the dark end the W sensel will be able to capture more photons,
but only on 1/4 of the area, so the benefit on the noise floor would
seem to be small.

It could be that the effects in the two paragraphs above cancel out so
no extra DR is gained.

We'd also lose 1/4 of the colour information. A colour image would
have to be recreated by using info from the remaining 75%. This
will increase croma noise a bit.

Finally, the splitting into G and W would seem to cause more moiré,
but at 48Mp the lenses might act as AA filters so this may not be a
problem.

Do you see any errors in my thinking?

Just my two oere
Erik from Sweden

 

[Plusiotis]

No thread.

 

[Paul Shields]

Interesting value judgement there to suggest that prints have more
value than a digitally displayed image. May be true, but to suggest
that digital images have no value until printed... will leave that
philosophical discussion for another day.

Well I'm a bit of a luddite who also thinks prints are the final arbiter. Showing a 32megapixel image on a screen at 72dpi seems quite pointless if that's the final destination. I like to make prints. If I just want to show images digitally then my ancient E-1 will more than suffice.

That aside 32MP works out to very close to 300 dpi on A3 and 150 dpi
on A1. Whether or not that is of use for you, I suspect there are a
commercially significant number of people for whom that level of
detail would suit just fine.

Quite agree - bigger prints with smaller dpi are the only reason for higher pixels.

--
my blog
http://photocamel.com/forum/blogs/paul-shields/

My pics
http://www.flickriver.com/photos/noisyoly/popular-interesting/

 

[Pete_CSCS]

It is currently impractical,

If by that, you mean it is uneconomical or that camera makers see
more profitable alternatives, then I agree.

In a sense, it doesn't really have to be practical to sell it.

The smaller exotic car manufacturers are continuously topping each other in producing high horse power, very fast and very expensive street legal automobiles.

Currently:
more than: $1,000,000.00 apiece
more than: 1000 - 1100 BHP
more than: 250 mph

Not very practical considering the speed limits, the price of a gallon of gasoline (very thirsty engines) and car insurance (collision would be through the roof) but does that stop them from being sold?

There is a waiting list for these cars (with sizable deposits laid down) before the first completed car rolls out of the factory!!

If Olympus built a 32-40 MP DSLR in small quantities, they would get swept up if nothing more than bragging rights from wealthy camera owners looking to show them off.

As others mentioned, the extra resolution could showcase the superiority of the digitally designed Zuiko ZD lenses. If the cameras sold better than expected, Oly could always increase production - a good situation to find yourself in.

Olympus seems to like to stick their neck out and not always play it safe - so why not?

--
Stay Well,
Pete K.

 

[DriftForge]

Well I'm a bit of a luddite who also thinks prints are the final
arbiter. Showing a 32megapixel image on a screen at 72dpi seems quite
pointless if that's the final destination. I like to make prints. If
I just want to show images digitally then my ancient E-1 will more
than suffice.

I had a discussion about this with my wife a few weeks ago. We came to the conclusion that printing added value to a image, but that there was value in images that weren't printed. I think the increase in value comes from 1) more work / cost being put into the printed image and 2) the additional accessibility.

And yes a screen portion is not much of 32M image..

 

[SirSeth]

We'll get to 32MP eventually, but I personally don't think the overall image quality will be much better than what we have now because hiking the MP has actually lost us quality that technology is desperately trying to recover and meet expectations of customers who have been trained by the industry and culture to buy based on stat sheets rather than real world performance and quality. So ha! My best run on sentence on this topic.

Cheers, Seth

--
What if the hokey pokey really is what it's all about?

--
wallygoots.smugmug.com
wallygoots.blogspot.com